New attenuation correction for the HRRT using transmission scatter correction and total variation regularization

IN the standard software for the Siemens HRRT PET scanner the most commonly used segmentation in the μ-map reconstruction for human brain scans is MAP-TR. Problems with bias in the lower cerebellum and pons in HRRT brain images have been reported. The main source of the problem is poor bone / soft tissue segmentation in these regions and the lack of scatter correction in the μ-map reconstruction. In this paper we describe and validate the new TXTV segmentation method (included in the HRRTU 1.0 and 1.1 user software) aimed at solving the bias problem. Its improvements fall in two parts: Firstly, it introduces the use of scatter correction, and secondly, it uses a simple threshold in combination with nonlinear total variation μ-map filtering. To validate the new transmission μ-map reconstruction with scatter correction and TXTV segmentation, we compared the average activity in 50 RoIs average on emission images reconstructed with the new TXTV and old MAP-TR μ-maps to images reconstructed with co-registered CT-based μ-maps on four 18F FDG scans. The CT μ-maps is considered a gold standard. The comparison show a significant decrease in the bias in cerebellum and pons, and a better agreement between images reconstructed with the new method and CT-based μ-maps.